对齐与安全论文趋势 2023–2026

RLHF、偏好优化、红队与安全对齐

统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 4,558 篇被归入该主题的论文。

主要结论

每年论文数

各会议论文数

按年份和会议统计

对齐与安全论文数，按年份和会议
年份NeurIPSICLRICMLCVPR合计
202313887101112438
2024348156246125875
20254554733001351,363
20264806235512281,882

代表论文

按年份和会议均匀挑选，方便快速看到每一届的样貌

全部对齐与安全论文

年份全部2023202420252026
会议全部NeurIPSICLRICMLCVPR

共 1,339 篇，第 17 / 27 页

  1. ICLR 2025Generative Classifiers Avoid Shortcut Solutions
  2. ICLR 2025Generative Monoculture in Large Language Models
  3. ICLR 2025Glimpse: Enabling White-Box Methods to Use Proprietary Models for Zero-Shot LLM-Generated Text Detection
  4. ICLR 2025Global Identifiability of Overcomplete Dictionary Learning via L1 and Volume Minimization
  5. ICLR 2025GNNs Getting ComFy: Community and Feature Similarity Guided Rewiring
  6. ICLR 2025Going Beyond Static: Understanding Shifts with Time-Series Attribution
  7. ICLR 2025GRAIN: Exact Graph Reconstruction from Gradients
  8. ICLR 2025Gramian Multimodal Representation Learning and Alignment
  9. ICLR 2025Group-robust Sample Reweighting for Subpopulation Shifts via Influence Functions
  10. ICLR 2025GSBA$^K$: $top$-$K$ Geometric Score-based Black-box Attack
  11. ICLR 2025Guaranteed Generation from Large Language Models
  12. ICLR 2025Gyrogroup Batch Normalization
  13. ICLR 2025h4rm3l: A Language for Composable Jailbreak Attack Synthesis
  14. ICLR 2025HADAMRNN: BINARY AND SPARSE TERNARY ORTHOGONAL RNNS
  15. ICLR 2025HarmAug: Effective Data Augmentation for Knowledge Distillation of Safety Guard Models
  16. ICLR 2025HASARD: A Benchmark for Vision-Based Safe Reinforcement Learning in Embodied Agents
  17. ICLR 2025Have the VLMs Lost Confidence? A Study of Sycophancy in VLMs
  18. ICLR 2025HelpSteer2-Preference: Complementing Ratings with Preferences
  19. ICLR 2025HERO: Human-Feedback Efficient Reinforcement Learning for Online Diffusion Model Finetuning
  20. ICLR 2025Hessian-Free Online Certified Unlearning
  21. ICLR 2025Higher-Order Graphon Neural Networks: Approximation and Cut Distance
  22. ICLR 2025How Does Vision-Language Adaptation Impact the Safety of Vision Language Models?
  23. ICLR 2025How Far Are We from True Unlearnability?
  24. ICLR 2025How much of my dataset did you use? Quantitative Data Usage Inference in Machine Learning
  25. ICLR 2025How to Evaluate Reward Models for RLHF
  26. ICLR 2025How to Verify Any (Reasonable) Distribution Property: Computationally Sound Argument Systems for Distributions
  27. ICLR 2025Human-Aligned Chess With a Bit of Search
  28. ICLR 2025HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere
  29. ICLR 2025I Can Hear You: Selective Robust Training for Deepfake Audio Detection
  30. ICLR 2025Improved Techniques for Optimization-Based Jailbreaking on Large Language Models
  31. ICLR 2025Improving Data Efficiency via Curating LLM-Driven Rating Systems
  32. ICLR 2025Improving Language Model Distillation through Hidden State Matching
  33. ICLR 2025Improving Long-Text Alignment for Text-to-Image Diffusion Models
  34. ICLR 2025Indirect Gradient Matching for Adversarial Robust Distillation
  35. ICLR 2025Infilling Score: A Pretraining Data Detection Algorithm for Large Language Models
  36. ICLR 2025Injecting Universal Jailbreak Backdoors into LLMs in Minutes
  37. ICLR 2025Inspection and Control of Self-Generated-Text Recognition Ability in Llama3-8b-Instruct
  38. ICLR 2025Instructional Segment Embedding: Improving LLM Safety with Instruction Hierarchy
  39. ICLR 2025Interpreting Language Reward Models via Contrastive Explanations
  40. ICLR 2025Inverse Constitutional AI: Compressing Preferences into Principles
  41. ICLR 2025Is In-Context Learning Sufficient for Instruction Following in LLMs?
  42. ICLR 2025Is Your Multimodal Language Model Oversensitive to Safe Queries?
  43. ICLR 2025Iterative Label Refinement Matters More than Preference Optimization under Weak Supervision
  44. ICLR 2025Iterative Nash Policy Optimization: Aligning LLMs with General Preferences via No-Regret Learning
  45. ICLR 2025Jailbreak Antidote: Runtime Safety-Utility Balance via Sparse Representation Adjustment in Large Language Models
  46. ICLR 2025Jailbreaking as a Reward Misspecification Problem
  47. ICLR 2025Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks
  48. ICLR 2025Joint Graph Rewiring and Feature Denoising via Spectral Resonance
  49. ICLR 2025Joint Reward and Policy Learning with Demonstrations and Human Feedback Improves Alignment
  50. ICLR 2025Judge Decoding: Faster Speculative Sampling Requires Going Beyond Model Alignment

在 Claude、Cursor 里直接问这个趋势

接上 Paperlist MCP 后，写综述、挑方向或准备投稿时，不用离开对话框就能拿到这页的数字：每年各会议的论文数、增长倍数、涨得最快的子方向和代表论文，AI 会按你的问题整理。

可以直接复制这句话问 AI

用 paperlist 查对齐与安全在 NeurIPS 2023–2026 的论文趋势，哪一年涨得最快？

AI 会调用的工具：get_topic_stats {"topic": "alignment-and-safety"}

两分钟接入 Paperlist MCP（Claude Code、Cursor、Codex、ChatGPT）

常见问题

2023–2026 年四大会议一共有多少篇对齐与安全论文？
共 4,558 篇，按年份分别是 2023 年 438 篇、2024 年 875 篇、2025 年 1,363 篇、2026 年 1,882 篇。
NeurIPS 2026 有多少篇对齐与安全论文？
NeurIPS 2026 有 480 篇对齐与安全论文，上一届是 455 篇。
ICLR 2026 有多少篇对齐与安全论文？
ICLR 2026 有 623 篇对齐与安全论文，上一届是 473 篇。
ICML 2026 有多少篇对齐与安全论文？
ICML 2026 有 551 篇对齐与安全论文，上一届是 300 篇。
CVPR 2026 有多少篇对齐与安全论文？
CVPR 2026 有 228 篇对齐与安全论文，上一届是 135 篇。
哪个会议的对齐与安全论文最多？
NeurIPS，2023–2026 年共 1,421 篇。

「对齐与负责任 AI」下的其他主题